Nozick's work is pornography for the critical thinker, a real meaty treat concerning political philosophy. Largely viewed as a response to John Rawls' A Theory of Justice, in fact Nozick responds to or extends any number of theories, though Rawls and his Theory of Justice do form a prominent part. Nozick lays out an argument that a minimal state (more than an ultraminimal state) is both moral and the limit of what could be considered an overarching moral state. That said, he spends part three of his book focused on a Utopian framework, which he describes as a minimal state permitting other subordinate experiments, to include those that are more intrusive, paternalistic, etc.
The first two parts of the book are dense and thorough. Nozick does some heavy thinking and works through problems to their natural outcomes, often to a degree beyond that of other more average authors. For the most part, these two parts are really excellent and spur the reader to do some real thinking of his or her own. There are some minor flaws, though. Nozick tends to reduce situations to a finite series of choices (often three), even though the reader might be able to think of not only additional possibilities, but even conceiving the problem in completely different frameworks and categories. The apparent arbitrariness of some of his arguments derived from choice multiple choice arrangements cause skepticism about their final validity.
The third part is a little disappointing after the strength of the first two. It almost feels like Nozick got tired and said, "To heck with it, that's close enough, they'll get the point," or some such. Not that there aren't some good points in there, but if he had stuck with his method in the first two parts, it would have both carried more weight and provoked deeper thoughts in the reader.
Nozick's assault on Anarchy and Anarchism is pretty brutal; having read at least some of those he mentions, i.e., Murray Rothbard, David Friedman, etc., none of their arguments hold up against Nozick's. Nozick's indictment of more statist solutions is not as comprehensive--that is, he really does not address it at every point it could be, and perhaps no one can really address both the vast number of possibilities in structure as well as the intentions and motivations behind them. However, he conclusively shows that any predetermined "just" final patterned distribution of goods/wealth is untenable; as soon as anyone attempted to spend, consumer, use, give away, lose, etc., any of the distributed wealth, the pattern would be broken and would require a still further redistribution to correct again.
Nozick admits the potential problem of past injustices being committed in the original acquisition or in the transfer of assets, but does not focus specifically on it given how ubiquitous the problem is. As Ludwig von Mises correctly realized in his Socialism: An Economic and Sociological Analysis:
All ownership derives from occupation and violence...That all rights derive from violence, all ownership from appropriation or robbery, we may freely admit to those who oppose ownership on considerations of natural law...This is the doctrine of natural law has called the war of all against all. The war ends when the actual relation is recognized as one worthy to be maintained. Out of violence emerges law. (pp 42-43)
Given that Nozick admits that "past injustices might be so great as to make necessary in the short run a more extensive state in order to rectify them," (p 231) and that Rawls and others are focusing on present situations as arising out of past injustices and requiring correction, this is a genuine weakness in Nozick's overall argument. That the redistribution of wealth is untenable remains, but leaves the moral issue festering.
Nozick confounds "risky" with intentionally malicious, i.e., talks about stealing property as a "risky" behavior, when in fact the risks are getting caught or failing to succeed in the larceny. A theft deliberately carried out isn't some risk like a person driving too fast in traffic who feels himself to be adequately in control--until he isn't. This slippery use of the language tends to undermine otherwise very thoughtfully and thoroughly constructed arguments. Nozick also largely dodges questions of restitution in cases where full restitution is not really possible--i.e., murder. He leaves Libertarian arguments about such things largely unexplored and unchallenged, though he cites some interesting works in his end notes. He casts a generally pessimistic view on deterrence, though he does not thoroughly explore if some level of deterrence is not a benefit worth some cost, even if most would agree no deterrence will be 100 percent effective in all cases--in any event, not all crimes are thoroughly thought through before commission.
Nozick's thinking and arguments jump around a lot and become a lot thinner in the Utopia part of the book (part three). He makes a lot of unsupported assumptions that seem counterfactual in the fact of real world history, probably engaging in the error of mirror-imaging:
No one will choose to be a queen bee...each person prefers being surrounded by a galaxy of persons of diverse excellence and talent equal to his own to the alternative of being the only shining light in a pool of relative mediocrity. All admire each other's individuality, basking in the full development in others of aspects and potentialities of themselves left relatively undeveloped. (p 306)
Like most Libertarians, Nozick skirts around the problem of children. If everything is by consent, what happens to any arrangement when children are brought into it, not through any initiative of their own, and not being asked nor agreeing to specific terms at birth or any time up to majority.

It really is a must-read, but is written at a level not accessible to all readers, so one has to be really interested in this topic to see it all the way through, though it is well worth it to one with such an interest.